By Bill Walsh, Industrial Hygiene Services Business Development, CIH
The new silica standard requires that analytical laboratories performing analyses under this standard must be accredited to ISO17025. SGS Galson is. We are also accredited for silica analysis by the American Industrial Hygiene Association (AIHA).
One of the hallmarks of the AIHA’s program is the requirement that laboratories participate in a proficiency testing (PT) program. Quarterly, the lab receives four samples and a blank spiked at unknown level (single blind). The samples are analyzed and the results submitted. The passing score is within three standard deviations of the reference level. If more than 25% of the samples lie outside this range, the laboratory fails the round. If a lab fails two consecutive rounds, they are considered non-proficient and cannot accept silica samples for analysis until they regain their good standing.
This program is a great way to compare the quality of a lab’s analysis relative to all other labs accredited for analyzing silica. Laboratories should be willing to share the results of their PT. If a lab will not provide this data, you should seriously question their ability to perform this analysis. Below is an example of SGS Galson’s PT results.
Testing Results for IHPAT Round 207 |
||||||||
Contaminant |
Units |
# |
Result |
Ref. Value |
Lower Limit |
Upper Limit |
z-Score |
Rating |
Silica |
mg |
1 |
0.0428 |
0.0517 |
0.0294 |
0.074 |
-1.2 |
A |
mg |
2 |
0.15 |
0.1705 |
0.1064 |
0.2345 |
-0.8 |
A |
|
mg |
3 |
0.0625 |
0.0711 |
0.0424 |
0.0998 |
-0.9 |
A |
|
mg |
4 |
0.1199 |
0.1329 |
0.0803 |
0.1854 |
-0.7 |
A |
The “Z-score” in column eight is the indication of pass or fail for a sample, which must be within plus or minus 3. Look at several rounds of data for your lab to get an idea of their accuracy and precision. Consistent outliers (outside 3 units) or passing samples with a score above 2 are indications of trouble.
Accreditation by itself should be considered the floor for evaluating a lab. As with everything else, quality in labs lies on a spectrum from good to bad. Clients must evaluate this quality or risk getting bad data.